Students who use AI generally score worse at school
OECD data shows students using AI for schoolwork often perform worse, but critical assessment training can reverse this trend.Summary
According to data from the OECD's 2025 PISA study, which assessed over 760,000 15-year-olds in 91 countries, students who use AI tools for schoolwork generally perform worse in science than those who do not. The negative impact varies by usage type: AI for specific tasks like summarizing or drafting writing correlates with worse performance, while broader use for research or general learning help shows a smaller drop. Frequency matters, with daily and very infrequent users performing worst. However, the study reveals crucial nuances. Students who used AI weekly for general learning support slightly outperformed non-users. This positive effect is significantly amplified when students are frequently trained to critically evaluate AI-generated information; in such cases, regular AI users outperform non-users. The authors suggest this indicates 'moderate and intentional use' of AI can be beneficial. OECD director Andreas Schleicher draws an analogy, stating technology helps when it enables 'productive cognitive struggle' but hinders learning when it 'short-circuits' this process. The study also notes AI use is more common among socioeconomically advantaged students and varies greatly by country, with rates over 95% in Vietnam versus about 60% in Japan. Interestingly, AI users showed higher curiosity levels, but this did not consistently translate to better grades.
(Source：The Verge)